Bleeding edge image salvage for newbie?

I have a RAW image taken with a Canon 400D Rebel where the area of interest is about a stop underexposed, far away, and not quite sharp. When I play with Levels to get normal skin tone it makes the ISO noise really bad. Tried turning off all sharpening (also seems to increase noise) in RawTherapee but wondering about tips for this worst-case scenario?

At what ISO did you take the image? You might have luck with noise reduction and/or LMMSE demosaicing in RT.

1 Like

Show your raw file http://filebin.net/
What is the intended viewing medium? An image on a webpage? A 90cm wide print?

ISO 800. I will start researching that method. For now, viewing on a computer screen is sufficient. This could end up as a print for legal purposes. Here is the download link …

RAW - group in center is important

I only had a few minutes to play. Here’s a quick take.

And the .pp3 I used: IMG_8514.CR2.pp3 (9.5 KB)

Wow; fast. Here’s what happens if I take an RT tiff into Gimp and lighten skin tones …

noise becomes visible

LOL spy photo.

w/o NR.

Yeah it’s the battle between levels, sharpening and noise that is the challenge.

1 Like

Funny, but in the case of television shows, there is ridiculous exaggeration of what kind of tools are available to law enforcement, and how well they work.

Theoretically, perhaps, but practically?

Before I wander off topic, Is the consensus opinion regarding my image that I am staring at the blank wall of reality??

As for the TV trope, my main beef is that I’ve met people who don’t understand why police can’t be more effective, considering “all the technology.” I think the media (entertainment or otherwise) is irresponsible, and I’m not prepared to blame the entire public for not being tech savvy.

Well, there’s no way to turn a Wurlitzer Spinet into a Steinway Grand.

1 Like

Furthermore regarding your image, it’s impossible to turn a 400D @ 55mm photo into a 5DSR @ 300mm photo.

I had just been loaned the camera by a friend, and was unfamiliar with it. This image was taken as I held the camera at about chest level and looked over the various controls and display. I squeezed the shutter in much the same way you might poke buttons and knobs inside a car you are considering purchasing. I was not trying to actually compose and “take a picture.” It wasn’t until later when three people in the image engaged in criminal mischief that the throwaway image became important.

But I do admit to a lack of familiarity with the software to process RAW images. I posted this question to see if I had overlooked a processing option. That appears not to be the main problem. Again, thank you all!

[quote=“okieman, post:12, topic:996”]
I think the media (entertainment or otherwise) is irresponsible, and I’m not prepared to blame the entire public for not being tech savvy.
[/quote]Yes it’s totally the medias fault if people can’t tell fact from fiction.

And now regarding recovering data from shitty images being impossible - that really depends on the constraints.

A quick example with a (simulated) license plate quite clearly blured beyond recognition.

Can we still find out which number it is? If we know the input image (the license plate) and the blur that was applied we can just apply the same blur and see if we get the same result.
Let’s assume it’s either 110 or 111:

Blur both in the same way:

Now let’s look at the differences between 110 and our unknown number:

Enhance! As they say:

Funky, looks like the last digit is not 0. Must be 111 then. :wink:

If you think that was a bit of a lame example maybe this is more convincing:

Now regarding the topic:
I doubt that you are going to get clearly visible faces out of that raw which I expect is what you want. If you had more images you could try some super resolution techniques but even then it’s probably going to be very hard.

1 Like

I hadn’t seen that type of image processing explained that well before; thank you for taking the time!

I only wanted to know if I had used good judgment in my general method-testing so far. For purposes of software exploration, this instance of using a low-quality image has been useful.